This paper undertakes a critique of experience curves from several angles. It considers the extent to which they can be regarded as an extension of learning curves, and concludes that the benefits from learning-by-doing at plant level are exhausted relatively early. It goes on to consider the evidence that there is a common slope to experience curves, their usefulness for forecasting prices, and possible reasons for a spurious correlation between accumulated output and average cost. It concludes by demonstrating the differences in strategic implications between the various possible economic factors which may underlie the experience curve. The conclusion is that experience curves are partly spurious, and of little practical value in forecasting or decision making.